Title :
Research on calculating method of hidden layer nodes in BP network
Author :
Shifeng Li ; Zhanzhi Qiu
Author_Institution :
Sch. of Mech. Eng., Dalian Jiaotong Univ., Dalian, China
Abstract :
According to the problem of time consuming in BP network hidden layer nodes number calculation by the traditional method, we propose the dimension projection algorithm. The hidden layer nodes calculation process of traditional method and genetic algorithm are analyzed, the dimension projection algorithm has been proposed, and the calculation process and parameter setting are amplified. We respectively use traditional method, genetic algorithm and dimension projection algorithm to calculate the number of single or double layer BP network hidden nodes of Iris dataset, Leaf dataset and 100 plant species leaves dataset. The simulation results show that the accuracy of dimension projection algorithm and traditional method are close, and the dimension projection algorithm greatly reduces the time consumption.
Keywords :
backpropagation; genetic algorithms; dimension projection algorithm; double layer BP network hidden nodes; genetic algorithm; hidden layer node calculating method; iris dataset; leaf dataset; plant species; Accuracy; Classification algorithms; Genetic algorithms; Iris; Optimization; Projection algorithms; Standards; BP Network; Genetic Algorithm and Dimension Projection Algorithm; Traditional Method; the Number of Hidden Layer Nodes;
Conference_Titel :
Control and Decision Conference (CCDC), 2015 27th Chinese
Conference_Location :
Qingdao
Print_ISBN :
978-1-4799-7016-2
DOI :
10.1109/CCDC.2015.7162423